Claude Fable 5 / 5.1
by Anthropic — the top generally available tier of the Claude family.
What it is
Fable is the publicly available version of Anthropic’s “Mythos-class” frontier model. Claude Mythos is the same underlying model with some safeguards removed, invite-only via Project Glasswing. Fable 5 ships with safety classifiers (cybersecurity, biology/chemistry, distillation) that can decline requests; per the launch announcement these route to Opus 4.8 and trigger in under 5% of sessions. Positioned for demanding reasoning and long-horizon agentic work; the advantage over Opus grows with task length and complexity.
Versions
| Fable 5.1 (current) | Fable 5 (legacy) | |
|---|---|---|
| Released | 2026-09-01 | 2026-06-09 |
| API ID | claude-fable-5-1 | claude-fable-5 |
| Knowledge cutoff | Jun 2026 | Jan 2026 |
| Retirement | not sooner than 2027-09-01 | not sooner than 2027-06-09 |
Shared specs: 1M-token context, 128K max output, token-based API pricing (Batch API 50% off), adaptive thinking always on (default effort high), text and image input, text output.
Fable 5.1 vs 5
Same base price, cache reads at a quarter of the previous cost, stronger long-running agentic coding, multistep research and document/spreadsheet/slide work. Breaking API changes: forced tool use returns an error; earlier models cannot read its thinking blocks; editing earlier turns invalidates thinking blocks. Additive (beta): per-message effort, turn-scoped system messages, progress updates between tool calls.
Availability
Claude API, Amazon Bedrock (anthropic.claude-fable-5-1), Google Cloud, Microsoft Foundry, Claude Platform on AWS; also claude.ai, Claude Code and Enterprise per the announcement.
Choosing
Anthropic’s docs suggest starting with claude-opus-5 for most workloads and moving to Fable 5.1 when evals on Opus 5.5 at higher effort still fall short.
Fable 5.1 announcement details (merged from the former Fable 5.1 note)
From Anthropic’s 2026-09-01 post (https://www.anthropic.com/claude-fable-and-mythos-5-1, fetched 2026-10-02):
- Cache reads cost 75% less; Anthropic puts the overall saving at about 25% for typical workloads and up to 45% for highly agentic ones.
- Claude Code users see around 60% fewer interventions per session from the cyber safeguards than on Fable 5; the biology safeguards fire 85% less often for benign elementary biology and medical questions.
- At lower effort settings Fable 5.1 is pitched as matching Fable 5 at much lower cost (earlier note; not re-checked).
- Enterprise Frontier Safeguards (EFS): detect and respond to misuse while giving enterprise customers the privacy of a zero data retention agreement; developed with “more than 100 customers”; rolling out in phases starting fall 2026. The earlier note also said data stays in the customer’s own cloud; that line was not found in the fetched excerpt (unverified).
- Companion release: Claude Mythos 5.1 (restricted access); see also claude-opus-5.
Sources
- https://platform.claude.com/docs/en/models/fable-5-1/overview (accessed 2026-09-30)
- https://platform.claude.com/docs/en/models/fable-5/overview (accessed 2026-09-30)
- https://www.anthropic.com/news/claude-fable-5-mythos-5 (accessed 2026-09-30)
- https://www.anthropic.com/claude-fable-and-mythos-5-1 (accessed 2026-09-30)
- https://platform.claude.com/docs/en/about-claude/models/overview (accessed 2026-09-30)
- https://platform.claude.com/docs/en/about-claude/model-deprecations (accessed 2026-09-30)